Samsung will collaborate with OpenAI to develop floating data centers and power plants as Sam Altman rushes to compete with his firm's own partners (www.techradar.com)

OpenAI and Samsung signed a letter of intent for a wide-ranging AI infrastructure partnership that covers high-performance memory supply, design and operation of AI data centers, and exploratory work on floating data centers and power plants. Samsung Electronics will be a strategic memory partner for OpenAI’s “Project Stargate,” with OpenAI projecting demand up to 900,000 DRAM wafers per month; Samsung SDS will help design, build and operate data centers and resell ChatGPT Enterprise in Korea; Samsung Heavy Industries and Samsung C&T will explore maritime data center and floating power/control-center designs. The companies highlight benefits such as easing land constraints, lowering cooling costs and cutting carbon emissions. Technically and strategically, the deal signals OpenAI’s shift toward vertical integration and operating scale rather than relying solely on hyperscalers (a move underscored by reports of massive Nvidia chip commitments). The memory supply commitment addresses a core hardware bottleneck for large models, while floating data centers introduce novel thermal, power and regulatory engineering trade-offs that could change deployment footprints—especially in coastal, land-constrained regions. For the AI/ML community, the partnership means tighter supply-chain alignment for model training, faster regional enterprise rollouts in Korea, and an acceleration of unconventional infrastructure designs that could influence latency, sustainability, and geopolitical distribution of compute capacity.